The Whole Life Of Rotating Machinery Condition Assessment And Diagnostic Studies

During operation of mechanical products, its health condition suffers from deterioration over time until functional failure occurs. Traditional condition monitoring and fault diagnosis simply define health condition of mechanical products as normal or failure. However, in practice, except for sudden functional failure, mechanical products suffer from a series of deteriorated conditions before functional failure. This process can be called as health deterioration process. In order to monitor different status during deterioration, such as mechanical early faults, and to prevent functional failure happening, health condition evaluation must be implemented to monitor mechanical current condition. After that, based on what have been observed about the past and current condition of system, proper maintenance decision can be made. In this process, early fault detection is very important because this is a detection point between mechanical system's normal and failure conditions. However, at the stage of early fault, fault related signal is very weak and prone to be overwhelmed by noise and other mechanical signals. Therefore, it is hard to detect when early fault occurs. Once early fault can be identified, maintenance decision-making can benefit from the result of health evaluation and machinery downtime can be reduced through proper decisions.Take gear and rolling element bearings as objects, research conducted in this thesis focus on mechanical product health evaluation, early fault detection and fault diagnosis.The main contents are given as follows:(1)Based on research on discrete wavelet theory, this thesis improves traditional combination of wavelet transform and Hilbert transform in the application of bearing faults diagnosis.(2) Based on wavelet modulus maxima theory, this thesis proposes a bearing fault diagnosis method through autocorrelation analysis of Lipschitz time-based function estimation.(3)Considering automatic bearing faults detection, Hidden Markov models are employed to distinguish different bearing faults.(4)This thesis proposes gear full lifetime condition evaluation and gear early fault detection methods on the basis of gear vibration mechanism.(5)By utilizing hidden Markov model and modern signal processing methods, this thesis proposes a method that can be used for monitoring gear health condition and detecting gear early faults.